Handle NULL bytes and control characters in source USFM

Our test team noted that when building a doc for English->Lingala (ln), the build fails with an error

This happened because the USFM for Lingala had control characters in
the text, in particular ^B (STX in ASCII - start of text). Since there
could be other source languages that have control characters I have
augmented STET to strip them out prior to processing.

backend/document/stet/stet.py

@@ -354,11 +354,14 @@ def generate_docx_document(
     #     template = filepath.read()
     env = jinja2.Environment(autoescape=True).from_string(template)
     full_html = env.render(data=word_entries)
+    # Remove any NULL bytes and all control characters
+    cleaned_full_html = re.sub(r"[\x00-\x1F]+", "", full_html)
     # logger.debug("full_html: %s", full_html)
     # filepath_ = f"{working_dir}/{lang0_code}_{lang1_code}_stet.html"
     filepath_ = f"{working_dir}/{document_request_key_}.html"
     with open(filepath_, "w", encoding="utf-8") as outfile2:
-        outfile2.write(full_html)
+        outfile2.write(cleaned_full_html)
+
     html_to_docx = HtmlToDocx()
     # docx_filepath = f"{Path(filepath_).stem}.docx"
     html_to_docx.parse_html_file(filepath_, f"{output_dir}/{Path(docx_filepath_).stem}")

tests/e2e/test_api_stet_docx.py

@@ -33,3 +33,18 @@ def test_en_abu_stet_docx() -> None:
             },
         )
         check_result(response, suffix="docx")
+
+
+@pytest.mark.stet
+@pytest.mark.docx
+def test_en_ln_stet_docx() -> None:
+    with TestClient(app=app, base_url=settings.api_test_url()) as client:
+        response = client.post(
+            "/stet/documents_stet_docx",
+            json={
+                "lang0_code": "en",
+                "lang1_code": "ln",
+                "email_address": settings.TO_EMAIL_ADDRESS,
+            },
+        )
+        check_result(response, suffix="docx")